今日内容

• Jupyter

  

  在此处摁下“m"键就会编程markdown模式
  

  摁”b“键是在下方创建代码

  

• %matplotlib inline

  作用：%matplotlib inline这一句是IPython的魔法函数，可以在IPython编译器里直接使用，作用是内嵌画图，省略掉plt.show()这一步，直接显示图像。

  如果不加这一句的话，我们在画图结束之后需要加上plt.show()才可以显示图像。不然就只有<matplotlib.collections.PathCollection at 0x2aa7c1ae490>这行话

  

  

一定导入的是matplotlib.pyplot 不然里面没有scatter方法，

scatter()方法的参数含义参考：https://blog.csdn.net/m0_37393514/article/details/81298503

• 具体代码实现如下

import pandas as pd

data = pd.read_csv("C:\\Users\\94823\\Desktop\\aaa.csv")

data

import matplotlib.pyplot as plt
%matplotlib inline

plt.scatter(data.Education,data.Income)

<matplotlib.collections.PathCollection at 0x2aa7c1ae490>

x = data.Education  # 定义横纵坐标
y = data.Income

import tensorflow as tf

model = tf.keras.Sequential()  # 建立一个输入一个输出的模型。比如图中的线性模型y=ax+b,x为一个输入，y为一个输出

model.add(tf.keras.layers.Dense(1,input_shape=(1,)))  
    # 前面的1对应的是units这个参数，表示该层有多少个神经单元。
    # 同时也是与上一层连接的W的列数。
    # input_shape的第一维一般表示一个batch的大小，
    # 即有多少个样本。后面表示每个样本应该有多少特征，也是多少个列。如果不写会根据输入的X自动推断。
    # 具体参考https://blog.csdn.net/weixin_42055967/article/details/104420985

model.summary()  # ax+b param中的两个参数就是a与b

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 1)                 2         
=================================================================
Total params: 2
Trainable params: 2
Non-trainable params: 0
_________________________________________________________________

model.compile(
    optimizer = 'adam',
    loss = 'mse'
)  # 减少误差优化的方法是梯度下降法：简写‘adam’，优化的损失函数是均方差‘mse’，
  # 通过这种方法编译

history = model.fit(x, y, epochs=5000)  # 把所有数据训练5000次

Epoch 4992/5000
1/1 [==============================] - 0s 3ms/step - loss: 287.1974
Epoch 4993/5000
1/1 [==============================] - 0s 2ms/step - loss: 286.9698
Epoch 4994/5000
1/1 [==============================] - 0s 3ms/step - loss: 286.7423
Epoch 4995/5000
1/1 [==============================] - 0s 3ms/step - loss: 286.5152
Epoch 4996/5000
1/1 [==============================] - 0s 2ms/step - loss: 286.2881
Epoch 4997/5000
1/1 [==============================] - 0s 2ms/step - loss: 286.0614
Epoch 4998/5000
1/1 [==============================] - 0s 3ms/step - loss: 285.8348
Epoch 4999/5000
1/1 [==============================] - 0s 3ms/step - loss: 285.6085
Epoch 5000/5000
1/1 [==============================] - 0s 3ms/step - loss: 285.3824

model.predict(x)  # 通过x预测

array([[25.934877],
       [26.824005],
       [27.802044],
       [28.691174],
       [29.58022 ],
       [30.558342],
       [30.473873],
       [32.336597],
       [33.225723],
       [34.203762],
       [35.092888],
       [36.20356 ],
       [36.96449 ],
       [37.849186],
       [38.73831 ],
       [39.716354],
       [40.60548 ],
       [41.494606],
       [42.47265 ],
       [43.361774],
       [44.250904],
       [45.228943],
       [46.118073],
       [47.007195],
       [47.89632 ],
       [48.874363],
       [49.763493],
       [50.674774],
       [51.630657],
       [52.519787]], dtype=float32)

model.predict(pd.Series([20]))

array([[48.088966]], dtype=float32)

明日计划

• 继续学习tensorflow